Effects of residency matching policies on regional distribution of residents: a simulation study using a stable matching algorithm

Abstract Background Appropriate geographical allocation of medical residents is often proposed to alleviate the regional maldistribution of physicians in Japan; however, its effectiveness remains unclear. We assessed the impact of adjusting residency positions and evaluated the effectiveness of different reallocation scenarios through simulation using a stable matching algorithm (the Gale-Shapley algorithm). Methods A cross-sectional simulation study was conducted using anonymized data from the Japan Residency Matching Program. A stable matching algorithm was implemented to simulate residency placement under multiple scenarios. Baseline scenarios adjusted the number of residency positions in each prefecture based on the physician maldistribution index at reduction rates of 5, 10, 15, and 20% (the 20% adjustment corresponded to a net reduction of approximately 1.5% in the total number of residency positions nationwide). Using the 20% reduction scenario, we further examined three reallocation scenarios: uniform reduction, reduction based on prior-year matching results, and reduction based on regional physician density. We compared the simulation results with the actual results of the residency matching in 2022. Results In the baseline scenarios, the matching process was simulated for 10,207 applicants. Fill rates (number of matched applicants divided by number of residency positions) declined slightly as adjustment rates increased, with the number of matched applicants ranging from 8,909 (82.6%) to 8,648 (81.0%), compared with 8,995 (82.9%) in the actual 2022 match. Despite this decline, minor changes were observed in rankings of the fill rates by prefecture or match rates (number of matched applicants divided by the number of applicants) by medical school, as indicated by the high Kendall’s τ coefficients (0.687–0.919). The three reallocation scenarios also showed no substantial changes from the 2022 results. The Gini coefficient for matched applicants remained unchanged compared with the 2022 results, indicating minimal variation across scenarios. Conclusion Adjusting residency positions based on regional maldistribution had a limited impact on reducing disparities. Regional disparities persisted even under alternative reallocation scenarios. Policy measures beyond quota adjustments, such as targeted admissions, nonmonetary incentives, and community engagement, may be more effective in addressing the maldistribution of residents.